Science in the Courtroom: EPA Glyphosate Edition

Countless articles at the American Council on Science and Health have reported on the studies underlying the regulatory science of glyphosate.  It would be fair to say that our analysis suggested little risk.  But we have also written numerous times about the need for regulatory transparency and the need for accurate scientific data underlying those decisions. In this instance, our belief in the need for consistent, transparent, and scientifically reliable processes means we are – brace for it, siding with the Rural Coalition. in this particular instance and calling for more data and a better, more holistic, or cohesive assessment.  The EPA failed its responsibilities and needs to redo its analysis to comport with basic scientific and legal standards.

Some background

Glyphosate, the active ingredient in Monsanto’s (and now Bayer’s) Roundup, is a very controversial chemical.  Some health agencies, like IARC, have found it to be a “probable human carcinogen,” although their analysis has been subject to pretty severe criticism.  Others, like the EPA, find it poses no “unreasonable risk” to humans.  Several court cases found the manufacturer liable for individual instances of cancer, resulting in damage awards that are being appealed.  Members of the agricultural community are also divided, with some fearful, and others touting its ability to reduce weeds and maximize yields.

Bayer AG, which purchased Roundup and its liability for $63 billion, hopes the litigation will end. To promote that vision, they are voluntarily removing Roundup from the consumer market, although it will remain available to agricultural users.  They have also agreed to pay close to $11 billion to settle many of the lawsuits already in motion which claim Roundup “caused” their cancers. The judge overlooking that settlement is not satisfied that the $2 billion in reserve payments is sufficient for future claims against their product.

An EPA statement that glyphosate poses no “unreasonable risk” would be helpful to Bayer in their appeals, deterring future cases and arriving at a settlement so the company is no longer distracted and drained by this litigation. Such an EPA finding would be especially welcome since the standard for EPA regulation is far more stringent than that required to determine causality in a civil tort action.

The Court’s Opinion

The Ninth Circuit, which evaluated the EPA administrative determination, began by reviewing the underlying process in the development of the EPA’s cancer assessment of glyphosate. The “EPA’s Cancer Paper, a generally applied, hazard assessment, constructs

“a total  analysis examining what the biological data reveal as a whole about carcinogenic effects and mode of action of the agent, and their implications for human hazard and dose response evaluation.”

In brief, the four-step process outlined in the 2005 Guidelines for Carcinogen Risk Assessment:

• Determines if a chemical can “present a carcinogenic hazard to humans,” and, if so, under what circumstances, utilizing a "Hazard Identification” process.
• Identifies at what level those adverse effects can occur and evaluates the dose-response
• Assesses the relevant conditions of human exposure, and
• Characterizes the risk – “including [h]ow well… data support conclusions about the nature and extent of the risk from various exposures.

The conclusion of this four-pronged analysis is reduced to one of five standard descriptors about human carcinogenic potential. [1] In the case of glyphosate, the EPA determined that it fit the category “Not likely to be Carcinogenic in Humans.”

Based on those 2005 Guidelines, that categorization is appropriate

“when the available data are considered robust for deciding that there is no basis for human hazard concern.” [emphasis added]

But the Court did not accept the EPA’s conclusion that the data were robust. The judges pointed to a blatant inconsistency in the EPA’s reasoning, a logical fallacy we face at the edges of our knowledge of cancer causation and the imprecise “science” of risk assessment. Commenting on an earlier finding, the Court commented:

“Because  inconsistent reasoning cannot survive substantial-evidence review, the [Scientific Advisory] panel concluded that EPA’s determination that glyphosate was not likely to be carcinogenic was not supported by substantial evidence.”

Absence of evidence is not evidence of absence

Indeed, the logical fallacy identified by the Court presents a valid concern. The EPA’s decision is based on their assessment of the totality of the evidence, including epidemiology and toxicology  (i.e., deriving from animal studies.). In both means of analyses, the EPA cherry-picked the data.

Regarding the animal  studies, they concluded that “animal tumor and genotoxicity studies showed no reason for concern.” Yet in their analysis in discussing the association between glyphosate and non-Hodgkin’s lymphoma (NHL), the EPA pointedly concluded that:

“a conclusion regarding the association between glyphosate exposure and the risk of NHL cannot be determined based on the available evidence.” [emphasis added]

If the risk cannot be determined, how can there be no reason for concern? It is this logical inconsistency that indelibly taints the remainder of their report. This inconsistency proves quite troublesome and detracts from the assurance that EPA provides that glyphosate isn’t a concern.

Selective Data – Cherry Picking

Specifically, the Court noted that the EPA used historical controls “only to discount studies indicating that glyphosate may cause tumors.” They also pointed out that the Cancer Guidelines provide two statistical methods to evaluate animal toxicity studies. In a pairwise test, one asks whether the treatment animals have more tumors than the controls. In a trend test, one asks if the tumor incidence increased as the dosage increased – was there a finding of a dose-response, a cardinal precept in finding a cancer association? 

According to the guidelines, failure of either test raises concerns of carcinogenicity.  But the EPA report discounts glyphosate’s positive trending data by relying on glyphosate’s negative pairwise data – a nice example of obtaining a different result when moving the goal posts.

The Court then notes other instances where the EPA may have been selective in its choice of epidemiological studies.  Thus, the EPA found and reported that:

• “Human exposure to glyphosate is associated with an at least somewhat increased risk of developing NHL” and
• “Some epidemiological studies provide evidence of an exposure-response relationship between glyphosate and NHL.”

But the EPA discounted only the adverse studies for a variety of reasons, however valid. These include excluding confounders, “such as exposure to other pesticides” driving the NHL results, or those indicating that the increased risk might be due to “chance and/or bias” (i.e., not statistically significant) or that some studies “did not detect a positive association between glyphosate exposure and NHL or an exposure-response relationship.”

We would argue that these are components of nuanced expert opinion and are not as troublesome as the logical fallacy (i.e., the inconsistency in their conclusions) stated both at the outset and conclusion and the cherry-picked results. Further, it is crucial to add that even the most compromising epidemiological readings (showing an increased risk of exposure with an elevated risk of cancer of 30 to 50%) would not serve to establish cancer causation in a civil tort action. Instead, the standards imposed on the EPA are based on the precautionary principle of prevention and are far more stringent than proving causation in a toxic tort matter.

Because “the EPA’s choice of a hazard description is not supported by substantial evidence,” among other reasons (including concerns raised by their own Scientific Advisory Panel), the Court vacated the EPA’s decision, giving them a “do-over” and told them to do it once again.   It’s entirely possible that they may come out with the same result on review, but at least they won’t be talking from both sides of their mouth, a far more reassuring assessment.

Let’s see what happens.

[1] The Guidelines provide five conclusions and their underlying criteria.  “Carcinogenic to Humans,” “Likely to Be Carcinogenic to Humans,” “Suggestive Evidence of Carcinogenic Potential,” “Inadequate Information to Assess Carcinogenic Potential,” and “Not Likely to Be Carcinogenic to Humans.”

Source: Ninth Circuit orders EPA to reexamine glyphosate’s toxicity to humans and the environment Courthouse News Service.